YundaDeskWhen a customer says “you sent me the wrong item,” this is not a normal support question. It is a trust problem. Move too slowly and the customer feels ignored. Move too fast and an agent may promise a replacement, refund, or compensation before the order and warehouse records are checked.
The better path is simple: own it fast, verify the order SKU and received item, then keep prepaid returns, replacements, and refunds under human approval. These scripts are useful for agents and for the knowledge base. AI can collect the facts; humans back up the risky parts.
Do not make the mistake sound smaller. Make the recovery cleaner: one less explanation from the customer, one less risky promise from the team.
– YundaDesk Support Team
Set the boundary: what AI can do
A wrong-item case can involve inventory, warehouse picking, logistics labels, after-sales policy, and cost. AI support can identify wrong item received, wrong color, wrong size, missing SKU, or mismatched item intent. It can check the order number, purchased SKU, received item, photos, product tag, barcode, shipping label, and outer packaging.
But prepaid return shipping, replacement, refund, compensation, and price change must be confirmed by a human. AI can prepare the summary and recommendation; it should not approve or execute. Refunds, compensation, and price changes always need approval and audit. If this line is not clear in your workflow yet, start here: AI answers first, humans back up.
Wrong-Item-Received Scripts: split handling permissions by risk first
Start with ownership, not interrogation
Do not begin with a long evidence demand. Do not explain warehouse operations first. Acknowledge the issue, then explain that you will verify it.
I am sorry the item you received does not match your order. I will first check the order and item details, then pass the case to a human teammate to confirm the right fix.
I am sorry this delivery did not go as expected. Please do not send the package back yet. I will first verify the SKU you ordered and the item you received, then we will confirm the next step.
Thanks for letting us know. To avoid asking you to ship to the wrong place or pay unnecessary postage, I need to check the order, item photos, and packaging details first. Anything involving replacement or return shipping will be confirmed by a human.
Avoid “we will send a replacement immediately,” “you can keep the wrong item,” or “we will fully refund you.” Those outcomes may be possible, but they should not be promised before verification.
Verify once: ask the right questions
Speed does not come from asking fewer questions. It comes from asking the right questions once.
| Information | Recommended script |
|---|---|
| Order number | Please send your order number or order email. I will check the product and SKU you purchased first. |
| Product photo | Please send a clear photo of the item you received so we can confirm the style, color, or size. |
| Product tag | If the item has a tag, barcode, size label, or SKU sticker, please send a clear photo of that too. |
| Packaging and label | Please keep the outer packaging and send a photo of the shipping label and package sticker, so our warehouse team can compare the records. |
| Item condition | Please confirm whether the item has been used, washed, opened, or whether the original packaging was discarded. |
Explain the fix: return, postage, replacement
The recovery step should answer three questions: do I need to return it, who pays postage, and when will the correct item be sent?
| Scenario | Script |
|---|---|
| Return may be required | After confirmation, if the wrong item needs to be returned, we will provide the return method. Please do not ship it back on your own yet, so the postage and address can be handled correctly. |
| Prepaid return shipping | If we confirm this was our fulfillment error, return shipping will be handled through the after-sales process. A human teammate will confirm the return method and postage coverage before sending instructions. |
| Return may not be required | For some low-value or special items, return may not be required. The team still needs to confirm based on the order, item condition, and local rules before we can say that. |
| Replacement item | We will first confirm the correct SKU and inventory status. A replacement must be created after human confirmation, and new tracking will be shared once it is arranged. |
| Customer needs it urgently | I understand you need the correct item as soon as possible. I will send the order, photos, and issue summary to a human teammate so the team can prioritize the workable fix. |
Use “we will confirm and arrange,” not “the system will handle it automatically.” Customers care about postage, the right item, and what they should do now.
Escalate high risk: refunds and compensation
Wrong-item cases often lead to refund requests, compensation claims, review threats, or marketplace complaints. This is where AI should stop in the right place.
I understand this mistake caused inconvenience. Refunds or compensation need human review based on order status, item condition, and after-sales policy. I will pass the order, photos, and conversation summary to the team.
If you prefer a refund instead of a replacement, we will record that request. Refunds are high-risk after-sales actions and can only be executed after human approval.
Avoid lines like “do not leave a bad review and we will compensate you,” “I will refund you now,” or “ship it back first and we will definitely reimburse postage.” They may calm the moment, but any later mismatch creates a second failure.
Keep channels consistent: no conflicting answers
A customer may start with the website widget, send photos by email, and then follow up on Instagram or WhatsApp. If those messages live in separate dashboards, agents ask for the same proof again or give conflicting answers.
Omnichannel support should bring the website widget, email, WhatsApp, Telegram, Messenger, Instagram, TikTok, LINE, WeChat, VKontakte, Zalo, YouTube, and custom API channels into one workspace and one customer profile. The agent should see the same order, issue, and recovery status. If your team is still cleaning this up, read: what an Omnichannel inbox actually solves.
For wrong-item cases, consistency matters more than raw speed. A fast answer that contradicts yesterday’s email makes the customer trust you less.
Feed the knowledge base: shorten the next case
After each wrong-item case closes, record four things: the original customer phrasing, the actual cause, the final resolution, and the wording that worked. For example: wrong item received, wrong size, wrong color, or received someone else’s item; warehouse picking error, SKU mapping issue, option misunderstanding, or logistics relabeling; prepaid return, replacement, refund, keep the item, or special approval.
Do not let AI learn this automatically. Use a controlled loop: when AI misses an answer, an agent replies, or an agent corrects AI, the system creates a learning suggestion. The owner reviews it before it becomes a skill, knowledge entry, or customer memory. Each change should be traceable, testable, and revertible. For the full approach, read: how to teach AI that gets smarter over time.
Appendix: knowledge base entry template for wrong-item cases
AI can handle directly
- Apologize and explain that the case will be verified
- Collect order number, product photo, tag, packaging, and shipping label
- Compare order SKU with the customer’s description
Must hand off
- Customer asks for refund, compensation, replacement, or price change
- Customer mentions complaint, review threat, or marketplace escalation
- Warehouse record, order SKU, and customer photos do not match
Approval rules
- Prepaid return shipping requires human confirmation
- Replacement requires inventory and correct SKU confirmation
- Refunds, compensation, and price changes always require human approval
The rule for wrong item received cases is simple: own it fast, verify it cleanly, and keep the fix under human control. Let AI collect facts and calm the first moment. Let humans approve replacements, returns, and refunds. Then fold the outcome back into the knowledge base, so the next case does not start from zero.